July centered on trust and control: finer-grained draft-review rules, better multi-channel orchestration, and expanded audit logging. Highlights below; watch the <a data-article-title="Breaking changes log" data-coll="Release Notes">breaking changes log</a> for any breaking changes and deprecations.
New
Per-tier and per-template review rules so you can auto-send low-risk tiers while keeping A-tier first-touches in review (see Review and approve agent drafts). Multi-channel orchestration improvements for sequencing LinkedIn touches between emails (see Multi-channel sequences). Custom roles for Enterprise expanded with more granular scopes (see Custom roles).
Improved
Audit log coverage widened to include more administrative and agent actions, with faster filtering (see Audit logs). Reply sentiment classification accuracy improved on short replies. Rate-limit headers now returned consistently on every endpoint (see Rate limits and quotas).
Fixed
Fixed a case where a healthy mailbox could briefly show Warming after a reconnection. Resolved a workflow pause that didn't surface its reason in the run log (see Why is my workflow paused?). Corrected an off-by-one in pagination cursors on large result sets (see Pagination patterns).
Notes
No breaking changes in July. If you consume the API, review the changelog for any upcoming deprecations before they take effect.
